RTX3080

[VBCurtis]

Originally Posted by chalsall

Please forgive me for this, but I'm "in that mode".

This is similar to the question "Can I cool a room by running a fridge with the door open?

Cold can't be created. Heat can be moved.

Unless there is a route to exhaust heat, a closed system will always get warmer.

Seems you are in "that mode", if "that" is where you preach about a topic that you not only don't know much about, but are mistaken.

Perhaps read a bit on evaporative cooling to see how the temperature of the room (sorry, "system") indeed decreases? In really dry places, swamp coolers work. Rather than educate us about their properties, try educating yourself.

[chalsall]

Originally Posted by VBCurtis

Perhaps read a bit on evaporative cooling to see how the temperature of the room (sorry, "system") indeed decreases? In really dry places, swamp coolers work. Rather than educate us about their properties, try educating yourself.

Always happy to be corrected.

Please tell me how evaporative cooling works without ejecting the heat from the system.

[ewmayer]

Originally Posted by chalsall

Please tell me how evaporative cooling works without ejecting the heat from the system.

Rather than spout multiple posts full of aggressive inanities as is your wont during these manic-troll episodes you are prone to, you might've asked yourself 'why did humans evolve sweat glands?' And rather than trot out technical terms you clearly don't understand like 'enthalpy', read up a little bit about the basics of phase transitions, featuring the term you clearly don't understand, the enthalpy of vaporization.

Better trolls, please.

[PhilF]

Originally Posted by VBCurtis

Seems you are in "that mode", if "that" is where you preach about a topic that you not only don't know much about, but are mistaken.

Perhaps read a bit on evaporative cooling to see how the temperature of the room (sorry, "system") indeed decreases? In really dry places, swamp coolers work. Rather than educate us about their properties, try educating yourself.

The key to Chris's post is "closed system". Swamp coolers work only if they are ingesting dry outside air, and if there's a way for that air to exit. If the cooler were just recirculating the inside air (meaning it's a closed system), then the cooler would not work.

[Uncwilly]

If one has an A/C system running, the system is not sealed in the house.
Evaporate the water inside (taking all of that heat to covert water to steam is 540 calories/g at 100C) using that device. Then the A/C unit not only cools the air, but condenses the moisture. That moisture is designed to go away. It may in the case of a window/wall unit, flow toward the outside. Some units actually are designed so the water gets picked up by the fan and flung against the condenser (to help transfer the heat to the air). Or in the case of a split system, it gets drained away into the sewer. Either way it gets moved out of the system of the air in the room.

[Cheetahgod]

If you put that air conditioning unit you bought outside then you can cool outside of your house creating a temperature differential causing the heat to leave your house.

[kriesel]

Originally Posted by ewmayer

In fact, I recently had to reroute the power cord to the desktop unit for precisely this reason - see the wall behind the desk & monitor in the pic? On the other side of that wall is the kitchen, and all the outlets there share the same 20A circuit as the dining room, whose corner the desk occupies - far right of the pic is the living room sofa, LR and DR form a large open 'L'. So as soon as I had > 1 GPU in the desktop unit up and running, every time someone turned on the eletric teakettle or the toaster (both short-run but high-draw appliances) plugged into the kitchen side of the shared wall, we'd trip the breaker.

I have systems scattered about the house, to make use of multiple breakers. All gpus that support it are set to somewhat reduced power via nvidia-smi or the graphical AMD equivalent. The worst performance hardware is downed for the summer.

Still have to be careful not to use too many appliances at once, and occasionally a breaker trips before I'm done vacuuming an area. And on a really hot day, the old air conditioner may not keep up with the 27 or 28C thermostat setting, without some manual load shedding intervention. Being in the midwest, summer outdoor humidity is generally uncomfortably high, precluding evaporative cooling.

[kriesel]

Originally Posted by chalsall

You are, of course, aware this doesn't change the enthalpy in a closed system?

It's not a closed system. That Ernst has not suffocated is proof. (And I'm very much in favor of that outcome.) Housing even with liberal application of window film to modern construction is pretty leaky. The leakiness is a necessary feature since the occupants are typically net oxygen consumers. (As are the gas water heater, gas stove/oven, and wood stove.) In the winter it takes several gallons water per day through humidifiers to maintain a healthy humidity level here (for human health and ESD mitigation) some of which is evaporated from 2 pots on the wood stove.

[kriesel]

Originally Posted by chalsall

Always happy to be corrected.

Please tell me how evaporative cooling works without ejecting the heat from the system.

Very similar to how ice cubes keep your drink cool for a while when you're watching the sunset with your SO; phase change to a different energy state, of part of the contents in an open volume. Meanwhile, the condensation on the outside of the container is the reverse of evaporative cooling.

[Xyzzy]

https://www.techpowerup.com/271284/p...dular-adapters

https://www.techpowerup.com/271257/n...-slot-behemoth

[ewmayer]

Originally Posted by PhilF

The key to Chris's post is "closed system". Swamp coolers work only if they are ingesting dry outside air, and if there's a way for that air to exit. If the cooler were just recirculating the inside air (meaning it's a closed system), then the cooler would not work.

Others have already addressed the incorrect closed-system assumption, but just by way of Gedankenexperiment let's imagine that the system were partially closed, in the sense that moisture could only enter by way of evaporation from the cooler, but had no way to leave once evaporated. I expect in a place the size of mine, somewhat over 10^4 cft (300 m^3), starting with typical CA summertime ambient air conditions meaning relative humidity ~= 40%, I could evaporate over 5 gal (say, 20 L) of water without the resulting relative humidity getting anywhere near 100% - perhaps someone has a set of suitable tables allowing this to be checked. Using Uncwilly's energy-of-evaporation number of 540 cal/g that translates to ~ 10^7 cal, or ~4*10^7 J. 1 J = 1 W*s, so 4e7 J ~= 11,000 W*h, roughly the same amount of heat put out by my 6-GPU compute setup in ~7 hours. [Any remaining cooling would need the AC, but 7 hours is pretty close to the typical windows-closed-and-AC-on time on a summer day here, i.e. this gets pretty close to obviating the need for AC cooling to offset the heat put out by the rig.]

Now in my above thought experiment, heat can still flow in and out of the space via windows and walls, so at night the cooler outside temps have a chance to remove enough heat to recondense the evaporated water - the only problem is that said condensation will be on every cool surface, collecting the water and returning it to the swamp cooler would be a problem. But aside from that, this describes more or less the same kind of moisture-sealed-in condensation loop used by your refrigerator and AC unit.

Anyhow, it'll be interesting to see how many gallons of H2O the unit I ordered actually manage to evaporate during a warm day, but the math indicates that each gallon evaporated offsets ~1.5 hours of the compute rig's 1.5kW heat production.

Getting back to the thread topic...

Originally Posted by Xyzzy

https://www.techpowerup.com/271284/p...dular-adapters

https://www.techpowerup.com/271257/n...-slot-behemoth

Holy crap, up to 600W per card? That would need the FLOPS equivalent of 3 Radeon VIIs running as sclk=3 to achieve parity on a FLOPS/W basis. But it'll be interesting to see the numbers once the 3080/3090 and R7 Pro start shipping and once someone gets an opportunity to run actual GIMPS-useful code on them.